SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Objective

Systemic lupus erythematosus (SLE) has been associated with an increased risk of cardiovascular disease. However, prospective population-based data addressing this association have been lacking.

Methods

We conducted a prospective cohort study among 119,332 women participating in the Nurses' Health Study who were free of cardiovascular disease and SLE at baseline in 1976. Incident SLE was confirmed by medical record review. Cardiovascular events included fatal and nonfatal myocardial infarction, stroke, coronary artery bypass grafting, and angioplasty. The relative risk (RR) of cardiovascular events among participants with SLE as compared with those without SLE was estimated using Cox proportional hazards models.

Results

Over 28 years of followup (2.9 million person-years), 8,169 cardiovascular events occurred and 148 women developed incident SLE. The mean age at SLE diagnosis was 52.6 years, and 20 participants with SLE developed a subsequent cardiovascular event. After adjusting for potential confounding factors, including age, race, cardiovascular risk factors, and medication use, the RR of a cardiovascular event in women with SLE compared with those without SLE was 2.26 (95% confidence interval [95% CI] 1.45–3.52). When end points were analyzed separately, the RR for coronary heart disease was 2.25 (95% CI 1.37–3.69) and the RR for stroke was 2.29 (95% CI 0.85–6.15).

Conclusion

In this prospective population-based study, we found a statistically significant >2-fold increased risk of cardiovascular disease among participants with SLE. The risk was not as high as has been previously reported, which may have been due to the relatively high age at diagnosis of SLE in this cohort.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Systemic lupus erythematosus (SLE) is a chronic autoimmune disorder occurring predominantly in women (1). Despite improved life expectancy in the past few decades (2), increased cardiovascular mortality and morbidity among SLE patients has been documented in several studies (3–7). Estimates of the magnitude of the increased cardiovascular disease risks among women with SLE range from a 50-fold increased risk of myocardial infarction among premenopausal women (3) to no increase in risk above that of the general population among elderly women (4, 7). This broad range in risk estimates may reflect the biology of cardiovascular disease among SLE patients, but may also reflect selection of SLE case populations and controls. Thus far, to our knowledge, no population-based study has prospectively examined the association between SLE and cardiovascular disease.

We investigated the association between SLE and incident cardiovascular disease among women participating in the Nurses' Health Study (NHS), the largest cohort of women followed prospectively for rheumatic disease.

SUBJECTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Study population.

The NHS is a cohort study of 121,700 female US nurses ages 30–55 years at the study inception in 1976 (8). The cohort was designed to prospectively examine relationships between lifestyle factors and chronic diseases, particularly cancer and cardiovascular disease. Information regarding diseases, lifestyle, and health practices is collected via biennial mailed questionnaires. More than 90% of the NHS participants have remained in active followup.

Identification of SLE.

As previously described (9), we employed a 2-stage procedure in which nurses were asked on the biennial questionnaires whether they had developed a physician-diagnosed connective tissue disease. Nurses who reported that they had such a diagnosis were sent a connective tissue disease screening questionnaire (response rate after 5 mailings 80%) (9, 10), and if this was positive, medical records were requested (response rate 74%) and reviewed. Two rheumatologists (EWK, KHC) trained in chart abstraction independently conducted a medical record review, examining the charts for the 11 American College of Rheumatology (ACR) classification criteria for SLE (11, 12). SLE cases were confirmed if they met ACR criteria (≥4 criteria) and if reviewers had consensus regarding definite SLE diagnosis. The case confirmation rate was 69% of the medical records reviewed and 7% of the original self-reports of any connective tissue disease (9), which is nearly identical to the self-reported rheumatoid arthritis confirmation rate of 6% reported for the Iowa Women's Health Study, another large, prospective, female cohort study (13). Furthermore, the rate of confirmed SLE in our cohort of ∼0.1% (9) is in line with recent estimates of the prevalence of SLE in the US (14).

Cardiovascular end points.

The end point for the current analyses was a composite end point (15) of fatal and nonfatal myocardial infarction, stroke, coronary artery bypass grafting, and angioplasty that occurred after the return of the 1976 questionnaire, but before May 31, 2004. We also investigated the outcomes separately. We requested permission to review medical records from women who reported having a nonfatal myocardial infarction or stroke on a followup questionnaire. Study physicians with no knowledge of risk factor status reviewed the medical records. Nonfatal myocardial infarction was confirmed if it met the criteria of the World Health Organization for symptoms plus either diagnostic electrocardiographic changes or elevated cardiac enzyme levels (14). Stroke was classified according to criteria established by the National Survey of Stroke, which required onset of a neurologic deficit with sudden or rapid onset that persisted for >24 hours or until death (15). We used self-reports for coronary artery bypass grafting and angioplasty because a validation in a subgroup indicated that self-reports from these nurse participants were virtually perfectly accurate.

Deaths were reported by the next of kin and the postal system or were ascertained through the National Death Index. Followup for the deaths was more than 98% complete in the NHS (16). Fatal myocardial infarctions and strokes were coded using the same criteria as nonfatal cases, but we additionally accepted autopsy evidence for both.

Covariates.

Information on diagnosis of hypertension, diabetes mellitus, and hypercholesterolemia was ascertained on each questionnaire. At baseline, participants reported parental history of coronary heart disease before age 60 years. Body mass index was computed for each 2-year time interval using the most recent weight in kilograms divided by the height in meters squared. Hours per week spent doing physical activities was assessed 7 times during the followup. Information on smoking status (never, past, or current) was assessed at baseline and updated every 2 years. Alcohol consumption was reported on the self-administered semiquantitative food frequency questionnaire (SFFQ) (17, 18). The baseline SFFQ was completed in 1980, and updated every 4 years. On each questionnaire, participants were asked about their menopause status and whether they used postmenopausal hormone therapy. From 1980 onward, information on use of aspirin was collected, and from 1990 and 1994 onward, information on use of nonsteroidal antiinflammatory drugs (NSAIDs) and oral corticosteroids was collected, respectively. In 1992, participants were asked to report their race, which was categorized as white, black, or other (Asian, American Indian, and Hawaiian).

Population for analysis.

At baseline, we excluded women with incomplete data on identification (n = 126), prevalent cardiovascular disease (n = 712), or connective tissue disease (n = 114). In addition, women who reported cardiovascular disease (n = 1,033) or connective tissue disease (n = 383) during the followup and had unconfirmed disease with the date of diagnosis missing were excluded at the start of the cohort. Therefore, 119,332 women were followed from 1976 until 2004. Women who reported a myocardial infarction or stroke that was not subsequently confirmed by medical record review were censored at the time of this report, as were women who reported any connective tissue disease other than confirmed SLE.

Statistical analysis.

We compared the presence of potential confounding factors and covariates among women diagnosed with SLE after the initial questionnaire in 1976 and the remainder of the cohort in 1990, the approximate midpoint of followup, as a representative time point (n = 103 women with incident SLE at that time). To prevent underrepresentation of cardiovascular disease risk factors, women already diagnosed with cardiovascular disease up to 1990 were included in these analyses.

We calculated the person-time of followup for each woman as the interval between the date of return of the 1976 questionnaire until the report of cardiovascular disease, death, or May 31, 2004, whichever came first. Women who developed SLE contributed exposed person-time from the questionnaire cycle following the date of confirmed SLE diagnosis. Relative risk (RR), the measure of association, was defined as the incidence rate of cardiovascular disease among women after the onset of SLE, divided by the incidence rate among women without SLE. We used Cox proportional hazards models to determine the independent association between SLE and the risk of incident cardiovascular disease while simultaneously adjusting for other risk factors. The Anderson-Gill data structure was used to efficiently handle time-varying covariates (19), where a new data record is created for every questionnaire cycle at which a participant is at risk, with covariates set to their values at the time that the questionnaire is returned. In these analyses, we used time-varying information from each 2-year interval to analyze the risk of cardiovascular disease in the next 2-year cycle.

Known or suspected risk factors for cardiovascular disease included in the multivariable proportional hazards models were hypertension, diabetes mellitus, hypercholesterolemia, parental history of coronary heart disease before age 60 years, body mass index (5 categories), physical exercise (hours per week), smoking status (never, past, or current), alcohol consumption (6 categories), menopause status and use of postmenopausal hormone therapy, aspirin, use of NSAIDs and oral corticosteroids (yes/no), and race (white, black, or other).

Because of limited power, we were not able to study the potential effect of SLE disease duration on the risk of cardiovascular disease through stratification by disease duration. Instead, we created a variable for disease duration using the baseline of each 2-year time period minus the year of diagnosis for women with SLE. Because SLE and SLE disease duration are highly correlated, we substituted SLE disease duration for SLE in separate regression models with cardiovascular disease as the dependent variable.

In our main analyses, only women with definite SLE contributed to exposed person-time (≥4 ACR criteria and reviewers' consensus regarding definite SLE diagnosis). The use of ACR criteria for diagnosing SLE biases against including milder or atypical cases or patients who do not fulfill criteria early in their disease (20), potentially underestimating the true incidence of the disease. Therefore, we examined the sensitivity of these analyses to the definition of SLE in secondary analyses using a broader definition of ≥3 ACR criteria and the reviewers' consensus on the diagnosis of SLE (probable SLE). Furthermore, because SLE may develop insidiously over several years, we conducted a sensitivity analysis in which we subtracted 4 years from the definitive SLE diagnosis date (which is synonymous with adding 4 years to exposure time) to potentially capture cardiovascular events that occurred close to but before the definitive diagnosis. All analyses were performed using SAS statistical software (SAS Institute, Cary, NC). The Partners Health Care System Institutional Review Board approved this study.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

In this study involving 119,332 women and 28 years of followup (2.9 million person-years), 8,169 cardiovascular events occurred and we confirmed a diagnosis of incident SLE in 148 women. At SLE diagnosis, the mean age of the women was 52.6 years and the mean number of ACR criteria was 4.8 (Table 1).

Table 1. Disease characteristics of the incident systemic lupus erythematosus (SLE) cases (n = 148) in the Nurses' Health Study at initial diagnosis*
 Value
  • *

    Values are the number (percentage) unless otherwise indicated. Manifestations of SLE are from medical record review. ACR = American College of Rheumatology.

  • Persistent proteinuria or cellular casts.

  • Hemolytic anemia, leukopenia, lymphopenia, or thrombocytopenia.

  • §

    Anti–double-stranded DNA antibody, anti-Sm nuclear antigen, and/or antiphospholipid antibody positive.

  • Four of 11 criteria are required for a diagnosis of SLE by ACR criteria.

Age at SLE diagnosis, mean ± SD years52.6 ± 8.6
Arthritis115 (78)
Serositis (pleuritis or pericarditis)52 (35)
Renal involvement28 (19)
Hematologic involvement72 (49)
Immunologic disorder§59 (40)
Antinuclear antibodies ≥1:40141 (95)
Number of ACR criteria, mean ± SD4.8 ± 1.0
Diagnosed by ACR member106 (72)

The distribution of potential cardiovascular risk factors among participants with SLE and those without SLE in 1990 is shown in Table 2. The mean ages were similar in the 2 groups. Women with SLE were more likely to have a history of hypertension and diabetes mellitus and to report a parental history of coronary heart disease before age 60 years. Participants with SLE exercised more, smoked less, but were more likely to have smoked in the past, and consumed less alcohol than those without SLE. Compared with women without SLE, women with SLE were more likely to be postmenopausal and to use postmenopausal hormone therapy. As expected, a much higher percentage of women with SLE than those without SLE reported regular use of steroids and NSAIDs. The majority of the participants were of white race, reflecting this cohort's racial composition, although a slightly higher proportion of the participants with SLE were of black race. In the middle of followup in 1990, the mean disease duration among women with SLE was 6.6 years.

Table 2. Age-standardized characteristics among participants in the Nurses' Health Study in 1990 according to history of validated incident systemic lupus erythematosus (SLE)*
 SLENo SLE
  • *

    Values are the percentage unless otherwise indicated. Variables were assessed at the 1990 questionnaire for participants who had previously been diagnosed with SLE (n = 103) and those without SLE (n = 108,968). NA = not applicable.

  • Data from 1994 are shown.

  • Asian, American Indian, and Hawaiian.

Age, mean ± SD years56.2 ± 6.756.2 ± 7.2
History of hypertension42.829.6
History of diabetes mellitus8.85.0
History of hypercholesterolemia34.034.2
Parental history of coronary heart disease before age 60 years22.215.6
Body mass index, age-adjusted mean kg/m225.425.6
Physical exercise, age-adjusted mean hours per week3.73.1
Cigarette use  
 Never44.143.2
 Past43.236.4
 Current12.416.9
Alcohol consumption, age-adjusted mean gm/day2.55.1
Postmenopausal73.766.6
Postmenopausal hormone therapy among postmenopausal women42.230.0
Aspirin use42.138.3
Nonsteroidal antiinflammatory drug use36.112.6
Steroid use29.81.3
Race  
 White95.196.8
 Black3.62.2
 Other1.31.0
SLE disease duration, age-adjusted mean years6.6NA

We calculated the age- and multivariable-adjusted RRs of the composite cardiovascular end point for women with SLE (Table 3). The age-adjusted risk of any cardiovascular event was >2-fold higher for participants with SLE than those without (RR 2.75, 95% confidence interval [95% CI] 1.77–4.27). After adjusting for potential confounding factors, the RR of any cardiovascular event for women with SLE decreased somewhat. When end points were analyzed separately, the multivariable-adjusted RR for coronary heart disease was 2.25 (95% CI 1.37–3.69). The number of incident strokes among the women with incident SLE was low (n = 4), and correspondingly the 95% CI for the RR was wide (RR 2.29, 95% CI 0.85–6.15). When coronary heart disease events were analyzed separately, the multivariable-adjusted RR for myocardial infarction was 1.81 (95% CI 0.75–4.37; n = 5) and the corresponding RR for coronary artery bypass grafting or angioplasty was 2.54 (95% CI 1.40–4.62; n = 11).

Table 3. Age- and multivariate-adjusted relative risks (RRs) for cardiovascular disease according to the presence of incident systemic lupus erythematosus (SLE) in the Nurses' Health Study, 1976–2004*
 SLENo SLEP
  • *

    RRs represent the hazard ratios computed using Cox proportional hazards regression analyses. 95% CI = 95% confidence interval.

  • Fatal and nonfatal myocardial infarction or stroke, coronary artery bypass grafting, or angioplasty.

  • Adjusted for age (years), history of hypertension, diabetes mellitus, and hypercholesterolemia, parental history of coronary heart disease before age 60 years, body mass index, physical exercise, smoking status, alcohol consumption, menopause status and use of hormone replacement therapy, aspirin, nonsteroidal antiinflammatory drugs and oral corticosteroids, and race.

  • §

    Fatal and nonfatal myocardial infarction, coronary artery bypass grafting, or angioplasty.

  • Fatal and nonfatal stroke.

Person-years of followup2,0822,932,407 
All cardiovascular end points   
 Incidence/100,000 person-years961278 
 No. of cases208,149 
 Age-adjusted RR (95% CI)2.75 (1.77–4.27)1.0 (ref)< 0.001
 Multivariable RR (95% CI)2.26 (1.45–3.52)1.0 (ref)< 0.001
Coronary heart disease§   
 Incidence/100,000 person-years769214 
 No. of cases166,284 
 Age-adjusted RR (95% CI)2.81 (1.72–4.59)1.0 (ref)< 0.001
 Multivariable RR (95% CI)2.25 (1.37–3.69)1.0 (ref)0.001
Stroke   
 Incidence/100,000 person-years19264 
 No. of cases41,865 
 Age-adjusted RR (95% CI)2.51 (0.94–6.69)1.0 (ref)0.07
 Multivariable RR (95% CI)2.29 (0.85–6.15)1.0 (ref)0.10

When studying the effect of disease duration, we found that a 1-year increase of SLE disease duration was associated with an RR of 1.08 (95% CI 1.04–1.11) for cardiovascular disease. This corresponds to an RR of 1.47 (1.085) for an SLE disease duration of 5 years. Multivariate adjustment did not materially change this risk (RR for 1-year disease duration 1.06, 95% CI 1.03–1.09).

Secondary analyses, using a broader definition of ≥3 ACR criteria and the reviewers' consensus on the diagnosis of SLE, included 152 women with probable SLE and produced results quantitatively similar to the main analyses (20 cardiovascular events; multivariable-adjusted RR for the composite cardiovascular end point 2.25, 95% CI 1.44–3.49). Starting SLE-exposed person-time 4 years prior to the date of SLE diagnosis from the medical records captured just one additional cardiovascular event (coronary artery bypass grafting/angioplasty) among the exposed, with a multivariable-adjusted RR for the composite cardiovascular end point of 2.14 (95% CI 1.39–3.29).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

In this prospective population-based study of 119,332 mainly white women with 28 years of followup, we found that women with SLE with a mean age at diagnosis of 53 years had a >2-fold increased risk of cardiovascular disease compared with women without SLE. Furthermore, we found that longer SLE disease duration was associated with an increased risk of cardiovascular disease.

The RR of cardiovascular disease in our study was lower than has been reported in several previous studies investigating the risk of cardiovascular disease among patients with SLE (3, 5, 6). Several factors may explain these different results. First, patients with SLE included in our cohort were generally older and were older at diagnosis than those included in previous studies. Because the absolute cardiovascular disease risk increases with age among all of the women, the relative cardiovascular disease risk among elderly SLE patients is proportionally lower. Furthermore, younger age at SLE onset is generally associated with more severe disease (21), which may be associated with a higher risk of cardiovascular disease among younger patients with SLE (22, 23). Second, unlike our study, none of the previous studies enrolled patients with SLE from the general population. Either patients with SLE from tertiary referral centers were included (3, 5) or hospital discharge data were employed (6), potentially leading to the selection of more severe SLE cases. Third, comparing cardiovascular disease rates among SLE cases from referral populations with cardiovascular disease rates among controls from the general population (3, 5) may have overestimated the magnitude of previously reported associations. Furthermore, the relatively high socioeconomic status of our population might have decreased the overall cardiovascular disease risk compared with other populations.

Our results are comparable with results from a population-based case–control analysis using general practice database data, which found a relative cardiovascular disease risk of 2 for women with SLE (7). When stratified by age, women in younger age groups were found to be at increased cardiovascular disease risk, whereas elderly women (age ≥70 years) were not (7). Also, in the Framingham Offspring Heart Study, in which a 50-fold increased risk of myocardial infarction among premenopausal women was reported, women ages 45–54 years had an RR of myocardial infarction of 2.47, which is comparable with our results (3). This decrease of risk with age most likely reflects the increase in absolute cardiovascular disease risk with age, but may also represent a selection phenomenon. Young women with SLE and vascular disease may have high mortality rates, and therefore not live long enough to become older women with cardiovascular disease (2). We restricted the analyses to incident SLE cases occurring after baseline in 1976, when the women in our study were ages 30–55 years, and followed the cohort for 28 years. Our results may therefore not apply to women with a younger age at SLE diagnosis. Although the mean age at SLE onset in this cohort was 53 years, these relatively older patients still had a >2-fold increased risk of cardiovascular disease. When end points were analyzed separately, the RR for coronary artery bypass grafting or angioplasty was somewhat higher than the RR for myocardial infarction. This might reflect increased cardiovascular surveillance among patients with SLE. However, the number of incident myocardial infarctions among patients with SLE in our population was limited and, correspondingly, the 95% CI for the RR was wide. Therefore, we cannot infer that results for coronary artery bypass grafting or angioplasty are significantly different from those for myocardial infarction.

The increased cardiovascular disease risk in patients with SLE is likely to be caused by a combination of factors (24). Traditional cardiovascular disease risk factors such as hypertension and diabetes mellitus are more prevalent among patients with SLE (25), but do not fully explain the risk in other studies (5), nor in our population. Whether SLE itself and/or its treatment increase cardiovascular disease risk is unclear. The inflammatory and immune mechanisms of SLE are considered to be of paramount importance in the pathogenesis of cardiovascular disease among patients with SLE (26–28). Additionally, procoagulant factors such as homocysteine and antiphospholipid antibodies may be involved in cardiovascular disease risk (29–31). Medications used to treat SLE may induce or protect from thrombotic events or atherogenesis. The widely held opinion that prolonged administration of corticosteroids accelerates atherosclerosis (32) has recently been challenged (22, 23). Furthermore, more frequent use of hydroxychloroquine and cyclophosphamide is associated with a lower risk of atherosclerosis (23), suggesting that control of disease activity with these medications might help prevent cardiovascular disease. Hydroxychloroquine may also favorably influence cardiovascular disease risk through lipid-lowering and antithrombotic effects (31, 33, 34). However, depending on the type used, both traditional NSAIDs and selective cyclooxygenase 2 inhibitors may increase the risk of cardiovascular disease (35, 36).

The strengths of our study include the large number of incident SLE cases, the prospective repeated assessment of exposures allowing for time-varying covariates, and the long duration of followup. We performed thorough medical record reviews in validating all of the cases of self-reported SLE, and only definite cases of SLE (≥4 ACR criteria) contributed to exposed person-time. Because the use of ACR criteria for diagnosing SLE biases against including milder or atypical cases or patients who do not fulfill criteria early in their disease (20), we examined the sensitivity of our analyses to the definition of SLE in secondary analyses using a definition of ≥3 ACR criteria and the reviewers' consensus on the diagnosis of SLE, and by beginning followup among SLE cases 4 years prior to the definitive SLE diagnosis date. These analyses produced results quantitatively similar to the main analyses, supporting the robustness of our results.

The results of our study must be interpreted within the limitations of the methodology used. Although the NHS is the largest population-based cohort of women who have been followed prospectively for rheumatic disease, the number of incident cardiovascular events among patients with SLE was limited. Also, any misclassification of SLE cases as noncases due to nonresponse to the request for information or insufficient data in medical records could have biased our study toward the null. However, the incidence rate of SLE in this study is comparable with rates reported in the literature, suggesting minimal misclassification. In our study, information on use of medication was limited to glucocorticoids and NSAIDs, with data collected since 1994 and 1990 only, respectively. Neither precise information on dose or duration of use nor data on use of other medications used to treat SLE was available. Because we did not review the updated medical records for each participant with SLE, we could not determine the cumulative disease severity over time. In addition, to maintain the prospective design, we excluded women who developed SLE at a younger age prior to the start of the cohort, who likely have more disease activity. Participants in the NHS are a select group of mainly white, educated women, and are not among the group at the highest risk of severe SLE; therefore, the generalizability of our findings to other racial and ethnic groups and those of lower socioeconomic status may be limited (37, 38).

In our prospective, population-based study, we found a >2-fold increased risk of cardiovascular disease among women with SLE. The risk was not as high as has been previously reported, which may be due to the relatively high age at diagnosis of SLE in this cohort. Furthermore, our findings support the importance of cardiovascular disease prevention and surveillance in SLE patients.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Dr. Hak had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Hak, Karlson, Stampfer, Costenbader.

Acquisition of data. Hak, Karlson, Costenbader.

Analysis and interpretation of data. Hak, Karlson, Feskanich, Costenbader.

Manuscript preparation. Hak, Karlson, Feskanich, Stampfer, Costenbader.

Statistical analysis. Hak, Feskanich, Costenbader.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

We gratefully acknowledge the participants in the NHS for their continuing cooperation.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. SUBJECTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES